|Publication number||US7369962 B2|
|Application number||US 11/030,728|
|Publication date||May 6, 2008|
|Filing date||Jan 6, 2005|
|Priority date||Jan 8, 2004|
|Also published as||EP1702242A2, US20050177326, US20080244371, WO2005065066A2, WO2005065066A3|
|Publication number||030728, 11030728, US 7369962 B2, US 7369962B2, US-B2-7369962, US7369962 B2, US7369962B2|
|Inventors||Torbjorn Sandstrom, Leif Odselius, Martin Olsson|
|Original Assignee||Micronic Laser Systems Ab|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (9), Non-Patent Citations (1), Referenced by (6), Classifications (7), Legal Events (8)|
|External Links: USPTO, USPTO Assignment, Espacenet|
This application is related to and claims the benefit commonly assigned of U.S. Provisional Patent Application Ser. No. 60/535,106, entitled “Method and Device for Data Integrity Checking”, filed on Jan. 8, 2004, naming Torbjörn Sandström, Leif Odselius and Martin Olsson as inventors. The related application is incorporated by reference.
The present invention relates to high speed datapaths, sometimes including mixed digital and analog voltage signals. In particular, it relates to error checking strategies for large data volumes, in digital and/or analog domains and to analog signal patterns that accelerate charge loading of micromirrors in an SLM.
A microlithographic direct writing system produces a very large data flow. Segments of the data path are not readily susceptible to error detection and correction, either because of the sheer data volume or because of the data format, such as an analogue output of a D/A converter. A direct writing system might include a positioning system for a workpiece, such as a wafer, on which a pattern is written. The writing begins with input pattern data that describes the pattern. A graphic data processing system with data storage and computing mechanisms converts transfer this data to a pattern exposure system that actually projects the image onto the workpiece. One direct writing system could be based on Micronic Laser System's SLM-based patterning platform for mask writing. An SLM is a spatial light modulator that includes a large array of small mirrors that are individually controlled by analog voltages output from D/A converters. A laser flash is relayed from the micromirror array to an image plane, where it exposes a pattern on the workpiece. When the flash is generated, the array of micromirrors is actuated in a state that represents an image element. The laser typically flashes at a rate of a few kilohertz. The number of micromirrors in an array is typically one to four million. The amount of data required to drive the micromirrors is very large and the required data transfer rate is very high.
It is difficult to preserve data reliability through the data-path to the exposure mechanism. Data reliability in this context means preserving data integrity so no pattern data information is lost or changed in a way that is not intended. The preservation of data integrity in the data path needs separate mechanisms applicable to digital data storage, digital data transfer, digital data manipulation, digital to analog data conversion and analog data transfer.
An opportunity arises to improve data reliability methods and mechanisms. Better, more trustworthy and reliable components and systems may result.
The present invention relates to high speed datapaths, sometimes including mixed digital and analog voltage signals. In particular, it relates to error checking strategies for large data volumes, in digital and/or analog domains and to analog signal patterns that accelerate charge loading of micromirrors in an SLM. Particular aspects of the present invention are described in the claims, specification and drawings.
The following detailed description is made with reference to the figures. Preferred embodiments are described to illustrate the present invention, not to limit its scope, which is defined by the claims. Those of ordinary skill in the art will recognize a variety of equivalent variations on the description that follows.
In an image processing system dealing with very large data volumes, the task of confirming reliability involves comparisons between different images and checking for discrepancies. Image processing systems in this context can be either based on image synthesis or image capture and analysis, or a combination.
A challenge found in a large-scale data set comparison is the ability to store and access the data that is to be compared. When data sizes exceed what is economical or practical to store on secondary storage media (such as hard-disk drives or tape) or when the requested rate of comparison exceeds what such devices exercise, the task of comparison between data sets become difficult. A byte-per-byte or block-per-block comparison would require fast access of all information to be compared. Many computation systems do not store calculation results for future use and the results only exist in an intermediate representation in data streams or volatile memory buffers. With the technology level of today (year 2003), data sets of multiple terabytes and access rates of multiple gigabyte per second represent performance levels where economical comparison can be hard to achieve.
The images used for comparison in an image processing system can be a combination of “known-good” data sets, representing a reference, and “to-be-verified” data sets representing a data set whose integrity one wishes to confirm. It is possible to use data sets solely of the “to-be-verified” type and compare them with each other and indicate discrepancies without stating which one is “correct”. It is also possible to use data sets of the “to-be-verified” in three or more instances and compare them with each other using a “majority decision” model. If two of the data sets match and a third set differs, a possible interpretation is that the third set is the one that is incorrect. It is important to note that the method so far is heavily dependent upon the definitions of what is “correct” and what is “incorrect”.
The methods proposed here address data volume and processing rate difficulties by using means of condensing or compaction to replace the task of comparing the entire data set with the task of comparing condensed or compacted versions of the data set. The method of condensation offers means of confining differences in the data set to specific regions, in which a more though comparison can be made. This way it is possible to trace individual error in a very large data set down to the level where the discrepancy can be exactly pointed out, in a two step procedure first resulting in a coarse pointer and then in a second step in an exact position. Locating errors without necessarily trying to correct them in real time is a more manageable problem.
A data set comparison method, such as the one described later in this document, can be applied to a plurality of situations where the data integrity is to be confirmed. The following are examples of situations where an efficient data set comparison method is useful.
When the physical execution components of a computation system of unknown reliability is to be verified with respect to data fidelity and repeatability (repeatability meaning the ability to produce identical results for each execution).
This verification can be requested at different situations such as (a) after production and assembly of the components, (b) after installation or repair of the components, (c) at regular intervals to check for long-term, permanent degradation of the components, or (d) during each execution of the computation to check for short-term or intermediate degradation of the components. It also can be requested when the software components of a computation system is to be verified with respect to data fidelity and repeatability, such as (e) introduction of new software versions that does not alter the data transformations, (f) introduction of new software versions that only partially alter the data transformations, (g) introduction of new software versions that completely alter the data transformations, or (h) investigation of in-field problems where the integrity of the software installation itself is under investigation.
The comparison method is also useful in a support situation where one wants to reproduce a particular execution at a different site with a different or identical setup. The proposed concept of comparison can then be used to ensure that the execution is reproduced in an identical way.
The proposed method prefers linear summation of data values over CRC, for condensed data sets. CRC (Cyclic Redundancy Check) uses the international standard CRC-32 algorithm. See, IEC 8802;ANSI X3.66. CRC methods give a very low probability that a two different data sets would give identical condensed value, and is hence the most secure to use. When the condensed value of two data sets differs, one can tell very little about the location of the difference, because CRC-32 algorithm is non-linear, so a one-bit difference in the data set can alter many or all bits of the condensed value.
Linear summation of data values, where the condensed value is the sum of values in the data set, gives a fairly high probability that two different data sets will generate the same condensed value, and can hence not be used to correct errors in the data or even to detect occasional two-bit errors. Consider for example a situation where two subsequent data values are swapped; in this case the linear sum would remain constant. However, linear summations have other more beneficial and useful characteristics, which will be shown later.
Partitioning of the Data Sets into Sections or Windows
One aspect of handling large data flow reliability is subdivision of the data set into sections or windows, as illustrated in
An extension of the concept of window division is shown in the
Condensing the Condensed Values
For very large data sets, the data sizes of condensed values in the map of sub-windows can grow relatively large. With a data set of 10 terabyte and for example 10.000 atomic data elements per sub-sub-window, the result is still 1 billion condensed value elements, which still can be difficult to efficiently transfer, access and compare. To resolve this without loosing too much of the precision in pinpointing individual discrepancies, the condensed values can be further condensed by performing an accumulation of condensed values over rows and columns of the sub-window mask, as illustrated in
By accumulating the condensed values of each Window Column Stack and each Window Row Stack, two new sets of matrices of condensed are created. One matrix set has a one matrix for each Window Column 412 and one matrix set has one matrix for each Window Row 422.
Monitoring Several Subsequent Processing Steps
Using Linear Summation for Algorithm Verification
An interesting property is that the linear summation, especially when the data set represents a chain of samples or pixels, reflects upon the average or aggregate magnitude of the samples.
In context where a particular data set is decomposed into sub-sections or sub-windows this representation offers a method of testing results of both transformation algorithms and algorithms used for the partitioning of data into the sub-sections or sub-windows. If, for example, changing the parameterization of such partitioning operation does not alter the individual data set values, but alters the linear summations on each window, one can check that the overall sum of all linear summations does not change with the parameterization change.
Using Condensed Values for Comparison in Production Systems
The concept described above facilitates efficient comparison with modest bandwidth and storage requirements. In a production system, the concept can be used to monitor the integrity of the data processing in near real-time. Four system configurations facilitating full or partial check of integrity are suggested here: 1) Ghost System; 2) Ghost Channels; 3) Using spare time between jobs; or 4) Using spare time during jobs.
This configuration, schematically illustrated in
The actual comparison is performed by a comparison process 930, separate from the processes that perform the processing. This process collects the condensed value data streams from both systems and matches them. The comparison process can execute on either a third, separate system, or on one of the compared systems. If the system latencies are short, the discrepancy report could potentially be used to initiate a system “retry” operation.
This configuration, schematically illustrated in
Using Spare Time Between Jobs
This configuration, schematically illustrated in
Using Spare Time During Jobs
Analog Data Integrity Verification
Data verification is substantially complicated when digital data is converted into an analog signal. Some additional terminology will help in understanding verification of analog signals used to drive micromirrors. The term static data integrity verification is used here to refer to data integrity when data is stored or transferred over a data link. A well-known method to verify static data integrity is the addition of redundant data, which is verified at the point where data is used. This redundant data can for example be data parity or cyclic redundancy check (CRC) or various types of checksums.
The term dynamic data integrity verification is used here to refer to data integrity when the data is manipulated, i.e. when data is changed by a computing element as required in graphic image processing. This can be done by addition of redundant computing elements, making simultaneous computations on the same data and comparing the results after each step. This approach has however disadvantages in that it requires much additional hardware, that it is difficult to analyze where computation malfunctions and to compare computations from one time to another.
Analog data integrity cannot be verified by addition of redundant data in a practical manner. A more practical method to get adequate verification of analog data transmission is to periodically verify the integrity of the data interconnect system and separately verify the signal response at a point somewhere along the analog signal line and as a last step verify the physical response from the analog signal. This is illustrated in
Assume that the analog source is a digital to analog converter 1402 with a buffer amplifier 1404. The analog output is monitored by a measurement device connected to point 1406. This measurement device might be an analog to digital converter 1408 that is connected to one of several signal sources by select switches 1410. The signal is further transferred to the signal destination which might be an spatial light modulator chip 1412. Means for monitoring the signal 1416 is attached to the signal as close as possible to the signal destination 1414. If the signal destination is an integrated circuit, this monitoring should preferably be integrated inside the circuit. The means for monitoring the signal 1416 should have an output signal that is connected to a monitoring device, which can be used to determine if the analog signal is connected to the signal destination. This monitoring is typically done periodically when the analog signal can be set to a known state but other configurations that allow continues monitoring is possible. The means for monitoring the signal from 1416 can be shared by several analog and digital input signals.
Another arrangement is shown in
The exclusive or gates in
Method to Transfer Analog Data at a High Sampling Rate
It is difficult to transfer analog data with a high speed via a sampled analog data link.
Some Particular Embodiments
The present invention may be practiced as a method or device adapted to practice the method. The methods can be viewed from the perspectives of digital or analog signals. Devices adapted to practice the claimed method are constructed from low level digital components that may be implemented in an FPGA, semi-custom ASIC or custom ASIC IC.
One embodiment is a method of detecting errors in a very high volume data stream to an SLM chip after generation of analog voltages to drive individual SLM elements. This method includes receiving one or more predicted diagnostic values that predict summary values for sets of analog voltages to be generated to drive the individual SLM elements. It further includes generating within the SLM chip one or more actual diagnostic values that summarize the sets of analog voltages being used to drive the individual SLM elements. It also includes comparing the predicted and actual diagnostic values and generating a signal corresponding to the comparison. The comparing may take place during or after generation of the actual diagnostic values. When it takes place during generation of the actual diagnostic values, it may operate on stored or pre-generated predicted diagnostic values. When stored actual diagnostic values are compared to predicted diagnostic values, the comparing may proceed during time slices within a job utilizing the SLM, when the SLM is not being used to generate patterns, or it may take place between jobs utilizing the SLM.
One aspect of this method is that the generating further includes switching among several sources of the analog voltages alternately coupled to a detector.
Another aspect of this method, applicable to any of the embodiments above, is that bi-level or multi-level detection may be applied to the analog voltages. Detection components may be calibrated using sample analog voltages to determine their actual response.
Any of the methods described above or aspects of the methods may be embodied in a device.
Another method embodiment is a method of detecting errors in a very high volume data stream using a summary value function. This method includes dividing the data stream into rows and columns of segments and calculating predicted diagnostic values that predict summary values for the rows and the columns of the segments. During operation of a data path that transmits the data stream to an SLM, this method includes calculating actual diagnostic values that summarize values for the rows and the columns of the segments. It further includes comparing the actual and predicted row and column summary diagnostic values and, in case of a discrepancy between the actual and predicted diagnostic values, identifying a segment causing the discrepancy from the row and column summary diagnostic value.
One aspect of this method is that the summary value function can be either a linear function or a nonlinear function. In the case of the linear summary value function, summing or repeatedly applying the linear summary value function across the row summary diagnostic values should equal summing or repeatedly applying the linear summary value function across the column summary diagnostic values and/or parts of the diagnostic values correspond to segments within the respective rows and columns. In the case of the nonlinear summary value function, summing or repeatedly applying the nonlinear summary value function across intermediate row summary diagnostic values should not equal summing or repeatedly applying the nonlinear summary value function across the column summary diagnostic values.
The aspects of the methods above, related to the timing or mechanics of comparing, may generally be combined with this method.
A further method embodiment is a method of high speed loading of charges that control numerous individual micromirrors of an SLM. This method includes selecting a particular voltage or charge level to control an orientation of an individual micromirror in the SLM and applying an analog voltage to a control line corresponding to the individual mirror, the analog voltage on the control line having a predetermined voltage or charge accumulation pattern that would exceed the selected particular voltage or charge level over an extended time. It further includes utilizing an accumulated voltage or charge at a predetermined time earlier than the extended time, the predetermined time being when the accumulated voltage or charge corresponds to the selected particular voltage to control orientation of the individual micromirror.
One aspect of this method further includes sampling the accumulated voltage on the control line at the predetermined time. An alternate aspect of this method includes accumulating a charge in a capacitor coupled to the control line and halting the charge accumulation at the predetermined time.
Any of the methods described above or aspects of the methods may be embodied in a device including logic and resources adapted to carry out the methods described.
While the present invention is disclosed by reference to the preferred embodiments and examples detailed above, it is understood that these examples are intended in an illustrative rather than in a limiting sense. It is contemplated that modifications and combinations will readily occur to those skilled in the art, which modifications and combinations will be within the spirit of the invention and the scope of the following claims.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US4547882||Mar 1, 1983||Oct 15, 1985||The Board Of Trustees Of The Leland Stanford Jr. University||Error detecting and correcting memories|
|US5691780 *||Jun 13, 1995||Nov 25, 1997||Texas Instruments Incorporated||Phase error control for color wheel|
|US5959747 *||Sep 11, 1996||Sep 28, 1999||California Institute Of Technology||Compact architecture for holographic systems|
|US6014128 *||Sep 27, 1996||Jan 11, 2000||Texas Instruments Incorporated||Determining optimal pulse width modulation patterns for spatial light modulator|
|US6151074 *||Sep 30, 1997||Nov 21, 2000||Texas Instruments Incorporated||Integrated MPEG decoder and image resizer for SLM-based digital display system|
|US6580490 *||Oct 30, 2000||Jun 17, 2003||Eastman Kodak Company||Method and apparatus for printing images in multiple formats using a spatial light modulator|
|US6618185||Nov 28, 2001||Sep 9, 2003||Micronic Laser Systems Ab||Defective pixel compensation method|
|US6697316 *||May 1, 2001||Feb 24, 2004||International Business Machines Corporation||Compensation of pixel misregistration in volume holographic data storage|
|US6958850 *||Oct 1, 2002||Oct 25, 2005||Corning Incorporated||Pointing angle control of electrostatic micro mirrors with modified sliding mode control algorithm for precision control|
|1||International Search Report of International Application No. PCT/SE 2005/000002, mailed Jul. 7, 20005.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US7793162 *||Aug 16, 2007||Sep 7, 2010||International Business Machines Corporation||Method for reporting the status and drill-down of a control application in an automated manufacturing environment|
|US7805639 *||Aug 16, 2007||Sep 28, 2010||International Business Machines Corporation||Tool to report the status and drill-down of an application in an automated manufacturing environment|
|US7835814||Aug 16, 2007||Nov 16, 2010||International Business Machines Corporation||Tool for reporting the status and drill-down of a control application in an automated manufacturing environment|
|US20080244371 *||May 5, 2008||Oct 2, 2008||Micronic Laser Systems Ab||Method and device for data integrity checking|
|US20090048701 *||Aug 16, 2007||Feb 19, 2009||International Business Machines Corporation (Ibm Corporation)||Method for reporting the status and drill-down of a control application in an automated manufacturing environment|
|US20090049011 *||Aug 16, 2007||Feb 19, 2009||Mock Michael W||Tool to report the status and drill-down of an application in an automated manufacturing environment|
|International Classification||G03F7/20, G06F11/08, H04N5/74, G06F11/00|
|Aug 15, 2007||AS||Assignment|
Owner name: MICRONIC LASER SYSTEMS AB, SWEDEN
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:SANDSTROM, TORBJORN;ODSELIUS, LEIF;OLSSON, MARTIN;REEL/FRAME:019696/0834;SIGNING DATES FROM 20050407 TO 20050412
|Nov 12, 2008||AS||Assignment|
Owner name: ASML NETHERLANDS B.V., NETHERLANDS
Free format text: LICENSE;ASSIGNOR:MICRONIC LASER SYSTEMS AB;REEL/FRAME:021817/0782
Effective date: 20041210
|Jun 17, 2009||AS||Assignment|
Owner name: ASML NETHERLANDS B.V., NETHERLANDS
Free format text: RELEASE OF LICENSE AGREEMENT;ASSIGNOR:MICRONIC LASER SYSTEMS AB;REEL/FRAME:022835/0217
Effective date: 20090612
|Jul 8, 2009||AS||Assignment|
Owner name: MICRONIC LASER SYSTEMS AB (PUBL), SWEDEN
Free format text: TERMINATION OF LICENSE AGREEMENT;ASSIGNOR:ASML NETHERLANDS B.V.;REEL/FRAME:022917/0810
Effective date: 20090707
|Sep 23, 2011||FPAY||Fee payment|
Year of fee payment: 4
|Dec 18, 2015||REMI||Maintenance fee reminder mailed|
|May 6, 2016||LAPS||Lapse for failure to pay maintenance fees|
|Jun 28, 2016||FP||Expired due to failure to pay maintenance fee|
Effective date: 20160506